Joydeep Chandra
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Joydeep Chandra has a total of 17 co-authors. Here are the top co-authors by number of shared publications:
- Satyam Kumar Navneet
Independent Researcher, Bihar, India
7 shared publications - Yong Zhang
BNRIST, Dept. of CST, Tsinghua University, Tsinghua University, Beijing, China
7 shared publications - Satyam Kumar Navneet
Department of CSE Chandigarh University Mohali India
6 shared publications - Prabal Manhas
Chandigarh University
3 shared publications - Rashi Sahay
Manav Rachna International Institute of Research and Studies
2 shared publications Aleksandr Algazinov
Dept. of CST Tsinghua University Beijing China
1 shared publication
Latest Publications
- Open Access
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
World Models: A Survey
World models, generative, spatio-temporal neural systems that compress multi-sensor observations into compact latent representations and simulate future outcomes under hypothetical actions, have emerged as a foundational paradigm for advancing artificial general intelligence. This survey provides the most comprehensive and critically analytical treatment of world models to date, spanning their conceptual origins in cognitive science and classical model-based reinforcement learning through to the cutting edge of modern neural architectures. We present a structured taxonomy of world model families including Recurrent State-Space Models (RSSMs), Transformer-based sequence world models, diffusion-based generative simulators, Joint Embedding Predictive Architectures (JEPAs), GAN-based world models, Neural Radiance Fields as implicit world models, State Space Models (Mamba-based), and multimodal hybrid systems that unify vision, language, and action. We systematically examine training methodologies including self-supervised and contrastive learning, reinforcement learning from human feedback, curriculum learning, physics-informed optimization, federated training strategies, and uncertainty quantification. A dedicated section analyzes the deep integration of natural language processing into world models, covering language understanding, procedural reasoning, dialogue grounding, retrieval-augmented generation, code-synthesizing world models, and cross-lingual capabilities. We survey applications across autonomous driving, robotics, time-series forecasting, scientific discovery, healthcare and biomedicine, financial modeling, social simulation, edge computing, and conversational AI. We introduce dedicated new sections on healthcare and biomedical world models, social world models and multi-agent simulation, financial and economic world models, and world models in edge and federated computing. We present a comprehensive Findings and Inferences section synthesizing cross-domain empirical observations, architectural insights, quantified limitations, and specific predictions for the trajectory of the field through 2030. This survey aims to serve as the definitive structured reference for researchers and practitioners across all areas of AI seeking to understand and advance the state of world models toward general intelligent systems.
- 29 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessWorld Models: A Survey
World models, generative, spatio-temporal neural systems that compress multi-sensor observations into compact latent representations and simulate future outcomes under hypothetical actions, have emerged as a foundational paradigm for advancing artificial general intelligence. This survey provides the most comprehensive and critically analytical treatment of world models to date, spanning their conceptual origins in cognitive science and classical model-based reinforcement learning through to the cutting edge of modern neural architectures. We present a structured taxonomy of world model families including Recurrent State-Space Models (RSSMs), Transformer-based sequence world models, diffusion-based generative simulators, Joint Embedding Predictive Architectures (JEPAs), GAN-based world models, Neural Radiance Fields as implicit world models, State Space Models (Mamba-based), and multimodal hybrid systems that unify vision, language, and action. We systematically examine training methodologies including self-supervised and contrastive learning, reinforcement learning from human feedback, curriculum learning, physics-informed optimization, federated training strategies, and uncertainty quantification. A dedicated section analyzes the deep integration of natural language processing into world models, covering language understanding, procedural reasoning, dialogue grounding, retrieval-augmented generation, code-synthesizing world models, and cross-lingual capabilities. We survey applications across autonomous driving, robotics, time-series forecasting, scientific discovery, healthcare and biomedicine, financial modeling, social simulation, edge computing, and conversational AI. We introduce dedicated new sections on healthcare and biomedical world models, social world models and multi-agent simulation, financial and economic world models, and world models in edge and federated computing. We present a comprehensive Findings and Inferences section synthesizing cross-domain empirical observations, architectural insights, quantified limitations, and specific predictions for the trajectory of the field through 2030. This survey aims to serve as the definitive structured reference for researchers and practitioners across all areas of AI seeking to understand and advance the state of world models toward general intelligent systems.
- Open Access
- 30 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 18-21
LEGEND: A Learned Explainable Graph-Enhanced Navigable Index for Hybrid Vector-Graph Search
We demonstrate LEGEND, a hybrid vector-graph index where knowledge graph structure guides HNSW construction, creating ''semantic shortcuts'' that improve navigation for graph-constrained queries. Unlike TigerVector and NaviX, which treat graph predicates as post-construction filters, LEGEND incorporates KG edges during neighbor selection, achieving 14-18% higher recall. Our system introduces KG-HNSW (reserving HNSW neighbor slots for KG-connected nodes), community-aligned partitioning via Louvain detection, attention-based score fusion with interpretable weight decompositions, and a learned query router selecting among four execution strategies. On FB15K-237 and a 100K-node Amazon product subgraph, LEGEND achieves 0.92 Recall@10 versus 0.78 for TigerVector, with ablations confirming KG-HNSW contributes the largest gain (+12%). Our demonstration enables attendees to visualize semantic shortcuts in real-time, inspect per-result score explanations, and stress-test recall under low-selectivity filters. A demo video is available at https://drive.google.com/file/d/1w1mN-jG69Wr0jOVCAaf_p26EmFhu6yc/view?usp=sharing [HERE].
- 30 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 18-21
Open AccessLEGEND: A Learned Explainable Graph-Enhanced Navigable Index for Hybrid Vector-Graph Search
We demonstrate LEGEND, a hybrid vector-graph index where knowledge graph structure guides HNSW construction, creating ''semantic shortcuts'' that improve navigation for graph-constrained queries. Unlike TigerVector and NaviX, which treat graph predicates as post-construction filters, LEGEND incorporates KG edges during neighbor selection, achieving 14-18% higher recall. Our system introduces KG-HNSW (reserving HNSW neighbor slots for KG-connected nodes), community-aligned partitioning via Louvain detection, attention-based score fusion with interpretable weight decompositions, and a learned query router selecting among four execution strategies. On FB15K-237 and a 100K-node Amazon product subgraph, LEGEND achieves 0.92 Recall@10 versus 0.78 for TigerVector, with ablations confirming KG-HNSW contributes the largest gain (+12%). Our demonstration enables attendees to visualize semantic shortcuts in real-time, inspect per-result score explanations, and stress-test recall under low-selectivity filters. A demo video is available at https://drive.google.com/file/d/1w1mN-jG69Wr0jOVCAaf_p26EmFhu6yc/view?usp=sharing [HERE].
- Open Access
- 28 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 1161-1170
TRACE: Transparent Web Reliability Assessment with Contextual Explanations
- 28 May 2026
- Published by Association for Computing Machinery (ACM)
- p. 1161-1170
Open AccessTRACE: Transparent Web Reliability Assessment with Contextual Explanations
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
When AI Writes, Whose Voice Remains? Quantifying Cultural Marker Erasure Across World English Varieties in Large Language Models
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessWhen AI Writes, Whose Voice Remains? Quantifying Cultural Marker Erasure Across World English Varieties in Large Language Models
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
The Topology of Recovery: Using Persistent Homology to Map Individual Mental Health Journeys in Online Communities
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-7
Open AccessThe Topology of Recovery: Using Persistent Homology to Map Individual Mental Health Journeys in Online Communities
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-8
TherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-8
Open AccessTherapyProbe: Generating Design Knowledge for Relational Safety in Mental Health Chatbots Through Adversarial Simulation
- Open Access
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-6
GazeFlow: Personalized Ambient Soundscape Generation for Passive Strabismus Self-Monitoring
- 13 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1-6
Open AccessGazeFlow: Personalized Ambient Soundscape Generation for Passive Strabismus Self-Monitoring
- 1 April 2026
- Published by Springer Nature
- p. 79-93
Real-Time Toxicity Detection and Mitigation in Social Media Using Reinforcement Learning
- 1 April 2026
- Published by Springer Nature
- p. 79-93
Real-Time Toxicity Detection and Mitigation in Social Media Using Reinforcement Learning
- 1 April 2026
- Published by Springer Nature
- p. 37-50
Aura-CAPTCHA: A Reinforcement Learning and GAN-Enhanced Multi-modal CAPTCHA System
- 1 April 2026
- Published by Springer Nature
- p. 37-50
Aura-CAPTCHA: A Reinforcement Learning and GAN-Enhanced Multi-modal CAPTCHA System
- Open Access
Orchestrating Attention: Bringing Harmony to the 'Chaos' of Neurodivergent Learning States
Adaptive learning systems optimize content delivery based on performance metrics but ignore the dynamic attention fluctuations that characterize neurodivergent learners. We present AttentionGuard, a framework that detects engagement-attention states from privacy-preserving behavioral signals and adapts interface elements accordingly. Our approach models four attention states derived from ADHD phenomenology and implements five novel UI adaptation patterns including bi-directional scaffolding that responds to both understimulation and overstimulation. We validate our detection model on the OULAD dataset, achieving 87.3% classification accuracy, and demonstrate correlation with clinical ADHD profiles through cross-validation on the HYPERAKTIV dataset. A Wizard-of-Oz study with 11 adults showing ADHD characteristics found significantly reduced cognitive load in the adaptive condition (NASA-TLX: 47.2 vs 62.8, Cohen's d=1.21, p=0.008) and improved comprehension (78.4% vs 61.2%, p=0.009). Concordance analysis showed 84% agreement between wizard decisions and automated classifier predictions, supporting deployment feasibility. The system is presented as an interactive demo where observers can inspect detected attention states, observe real-time UI adaptations, and compare automated decisions with human-in-the-loop overrides. We contribute empirically validated UI patterns for attention-adaptive interfaces and evidence that behavioral attention detection can meaningfully support neurodivergent learning experiences.
Open AccessOrchestrating Attention: Bringing Harmony to the 'Chaos' of Neurodivergent Learning States
Adaptive learning systems optimize content delivery based on performance metrics but ignore the dynamic attention fluctuations that characterize neurodivergent learners. We present AttentionGuard, a framework that detects engagement-attention states from privacy-preserving behavioral signals and adapts interface elements accordingly. Our approach models four attention states derived from ADHD phenomenology and implements five novel UI adaptation patterns including bi-directional scaffolding that responds to both understimulation and overstimulation. We validate our detection model on the OULAD dataset, achieving 87.3% classification accuracy, and demonstrate correlation with clinical ADHD profiles through cross-validation on the HYPERAKTIV dataset. A Wizard-of-Oz study with 11 adults showing ADHD characteristics found significantly reduced cognitive load in the adaptive condition (NASA-TLX: 47.2 vs 62.8, Cohen's d=1.21, p=0.008) and improved comprehension (78.4% vs 61.2%, p=0.009). Concordance analysis showed 84% agreement between wizard decisions and automated classifier predictions, supporting deployment feasibility. The system is presented as an interactive demo where observers can inspect detected attention states, observe real-time UI adaptations, and compare automated decisions with human-in-the-loop overrides. We contribute empirically validated UI patterns for attention-adaptive interfaces and evidence that behavioral attention detection can meaningfully support neurodivergent learning experiences.